According to a research note from CLSA, the global technology sector's AI supply chain is once again being troubled by concerns over politics, data centers, funding, and the Anthropic IPO, but the firm believes these fears appear exaggerated and analyzes them along four main lines: politics, gigawatt capacity, trillion-dollar funding needs, and IPOs.

On politics, the firm notes that both parties have concerns about data centers, but most lean toward setting guardrails rather than a full moratorium, offering a relative breather. About 150 gigawatts of data centers are under construction in the United States, with another roughly 46 gigawatts outside the US and China, totaling about 200 gigawatts. While the market worries there will be nowhere to put the chips, the firm instead believes chips may be in short supply. From now to 2030, incremental AI-related demand may be about 110 gigawatts, and some delays could help balance overly aggressive data center plans against Taiwan Semiconductor Manufacturing Co Ltd (NYSE: TSM) capacity.

On funding, the firm estimates that to achieve mid-teens capital returns, cloud computing revenue may need to rise from about US$450 billion in 2026 to about US$3.5 trillion by 2035, equivalent to shifting only about 175 basis points of global GDP. Compared with past peaks in computerization and telecommunications end-user spending, which respectively reached about 4.5% and a combined roughly 7.2% of global GDP, the claim that "the money will run out" is seen as excessive.

The firm also notes that the Anthropic IPO is stirring sentiment, but it estimates Anthropic accounts for about 11% of global tokens excluding China, while OpenAI accounts for about 27%. The market appears to be spending 100% of its energy worrying about a player with about a one-tenth share. It also lists more than 100 new AI labs with combined private valuations exceeding US$650 billion, showing the industry landscape is still unsettled.